ElevenLabs Music v2: genre switches mid-track and section-level editing
ElevenLabs released Music v2 on May 26, 2026. The model can switch genres within a single track, regenerate any selected section without touching the rest, and embed sound effects inside songs. It is trained on licensed data, cleared for commercial use, and arrived alongside price cuts of up to 50 percent.
The AI music race got its next sprint. ElevenLabs shipped Music v2 on May 26, and the feature list reads like it was written by someone who has actually sat in a session: genre transitions inside a single track, opera into heavy metal and back if you want it, fast dense vocal delivery that holds together, and sound effects embedded directly into the music.
The production-relevant piece is the editing. Improved inpainting means you can select any section of a generated track and redo just that part. The bridge is weak? Regenerate the bridge. The chorus stays untouched. Tracks can be assembled section by section, which moves these tools from slot machine toward instrument.
The licensing posture matters as much as the features. The model is trained only on licensed data and its output is cleared for commercial use, with Decrypt reporting partnerships across Believe, Kobalt and Merlin. ElevenLabs also cut prices alongside the launch, up to 50 percent on the API and up to 40 percent for self-serve creative customers, and reportedly crossed $500 million in annual recurring revenue in April.
Google's Lyria 3 Pro, Suno v5.5 and Stability's Stable Audio 3.0 are all circling the same market. For working producers, the takeaway is that legally clean AI audio is getting better and cheaper at the same time, and the editing workflows are starting to respect how music is actually made.